Dam-based Evolutionary Image Segmentation Using Quality Function and Union-Find Set

This paper explores the use of an evolutionary approach in the context of image segmentation to overcome the problem of specifying manually the number of clusters with the normalized cut approach. The proposed approach uses a quality function, a dam-based representation, and an Union-Find Set decoding method. The quality function provides an unbiased criterion and the dam-based representation can shorten chromosomes. The approach first splits raw images to a set of small homogeneous basins separated by dams, and then maximizes the quality function by dam-based genetic algorithm. The satisfactory experimental results on color images are obtained